Research On Multi-dimensional And Massive Social Network Visualization Technology

With the rapid development of information technology, the social network data which contains multidimensional and massive characters shows an explosive growth. Research on social network data visualization technologies has important significance.On base of analyzing main problems of multidimensional and massive visualization technologies, research in this thesis mainly focuses on the community discovery algorithm, force-directed algorithm in the massive data visualization technologies and the attributed graph clustering algorithm in the multidimensional data visualization technologies.Firstly, the existing community discovery algorithms are inefficient and can't satisfy the requirements of graph visualization. The force-directed algorithms have the problem of low efficiency and less obvious community structure. In order to solve these problems, a community discovery algorithm for massive social network data and a community layout algorithm for showing large-scale social network community structure are proposed. First, a community discovery algorithm which is based on Louvain algorithm for large-scale social networks is proposed. This improved algorithm combines scale-free and small-world features of social networking. And by using the method of selecting seed nodes in advance, the improved algorithm can inhibit the excessive merger of the large community in the Louvain algorithm, and merges small community timely. Second, a community layout algorithm for displaying community structure of massive social networking is proposed. By using community gravitational force, the algorithm can prompt nodes of the same community together, optimize community gravity model, simplify the algorithm steps. The experimental results show that the above algorithms can show the massive social network data clearly and effectively.Secondly, the existing attributed graph clustering algorithms are inefficiency and can't satisfy the requirements of graph visualization for the community quality. These algorithms also have problems of dimension disaster and human intervention. Aiming at the problems above, an improved attributed graph clustering algorithm and a multidimensional data visualization algorithm based on attribute mapping are proposed. The improved attributed graph clustering algorithm uses PCA(Principal Component Analysis) and SOM(Self-Organizing Map) to reduce the dimension of data and complete the clustering. Then based on the clustering results, the algorithm uses the attribute similarity formula to convert the original non-weight network graph to weighted network graph. In the end, communities are divided by using the improved community discovery proposed in this thesis to get the final results. The multidimensional data visualization algorithm which based on attribute mapping uses parallel coordinates to show data dimensional information through the view of community. The above algorithms are adaptive and non-supervision and they can meet the requirements of large-scale social network for community partition results.Thirdly, a visualization scheme of multidimensional and massive social network data is proposed by integrating above improved algorithms. And a visual prototype system is developed at the same time.